Skip to main content

Factor VI

Evidence by Default

Principle

Every sensitive activity should produce evidence as a normal consequence of performing work.

Evidence is not created for auditors. It is a first-class architectural output of governed systems, allowing decisions to be explained, posture to be verified and governance to be trusted.

Problem

Governance cannot be verified without evidence.

Too many organisations generate evidence only when an audit approaches: screenshots are collected, spreadsheets completed and approvals reconstructed from emails or chat messages. These artefacts describe what people believe happened rather than what the system actually recorded.

Governed work should produce its own receipts. Every sensitive activity should automatically emit the evidence needed to explain what happened, who performed it, which governance applied and why the outcome was considered acceptable.

Characteristics

Evidence is produced automatically

Evidence is emitted as a normal consequence of performing governed work—not assembled later for an audit.

Evidence explains activities

Evidence should explain what happened, who performed it, when it occurred, which governance applied and why the outcome was permitted.

Evidence is attributable

Evidence should be machine-readable, attributable and tamper-evident so it can be independently verified.

Evidence is non-repudiable

Evidence should be sufficiently trustworthy that participants cannot credibly deny the activity occurred or dispute the governance under which it was performed. Digital signatures, trusted timestamps, provenance and tamper-evident logs all strengthen non-repudiation.

Digital signatures, software attestations, SLSA provenance, transparency logs and trusted timestamps are all mechanisms for strengthening non-repudiation.

Evidence is reusable

The same evidence should support operational dashboards, incident response, compliance reporting and formal audits without being recreated.

Positive examples

Mortgage approval receipt

Approval, decision, approver, policy version and customer identifiers are recorded automatically when the mortgage is approved.

Release provenance

Build provenance, SBOMs and signatures are produced automatically as part of creating a software release.

Payment authorisation log

Every payment records the governance decision, obligations satisfied and resulting transaction identifiers.

AI tool invocation

Each tool invocation records the requesting agent, governing decision, data accessed and resulting actions.

Negative examples

Screenshot Evidence

A compliance team assembles screenshots before an audit to demonstrate that controls are configured correctly. The screenshots cannot explain what actually happened or be regenerated from the governed system.

Spreadsheet Attestations

Control owners complete quarterly spreadsheets declaring systems to be compliant. The evidence is disconnected from the activities it claims to describe.

Chat-Based Approvals

Production releases are approved in a chat channel without generating a durable governance record. Future reviewers cannot determine exactly what was approved or under which governance.

Anti-patterns

Audit Theatre

Evidence is created for auditors instead of being produced by the governed system.

Evidence by Exception

Receipts are generated only for unusual events rather than every sensitive activity.

Unattributable Evidence

Logs and reports cannot be traced back to a specific activity, governance version or responsible identity.

Disposable Evidence

Operational telemetry, provenance and decision logs are discarded once the activity completes, preventing future verification.

Diagram

How do we prove compliance?

Evidence is emitted as the activity happens, then consumed later by auditors and tooling—not reconstructed afterwards.

Discussion

Every sensitive activity should produce two outputs: a business outcome and an evidence receipt.

The business outcome is why the activity was performed—a mortgage is approved, a payment is transferred, a software release is deployed. The evidence receipt explains how that outcome came about by recording the implementation version, governance version, decision, identity, timestamp and resulting outcome. Rather than embedding governance definitions themselves, the receipt references the versions that governed the activity, allowing the complete context to be reconstructed later.

Evidence should therefore be treated as a first-class architectural output rather than an administrative afterthought. It should be emitted automatically by the governed system as the activity occurs, not assembled later for compliance exercises or audits. If evidence has to be reconstructed from screenshots, spreadsheets or chat logs, the system has already failed this principle.

Applying this principle means:

  1. Treat every sensitive activity as producing an evidence receipt.
  2. Record immutable references to the Implementation Twin, Governance Twin, governing decision and the business context needed to explain the activity.
  3. Include sufficient information to establish attribution, integrity and non-repudiation.
  4. Allow multiple governance functions to consume the same evidence, including operations, security, verification and audit.

This pattern already appears in many successful governance technologies. SLSA provenance, in-toto attestations, software signatures, transparency logs, OpenTelemetry traces and audit logs all generate evidence as part of normal operation. The Governance Factors extend this idea beyond software supply chains to every governed sensitive activity.

References

  • SLSA — provenance as normal build exhaust
  • in-toto — signed links for authorised steps
  • CycloneDX — machine-readable composition evidence
  • OSCAL — portable assessment results
  • OpenSSF Scorecard — continuous repository posture signals
  • Gemara — semantics for binding evidence to governance layers